Journal of Economic Literature 2010, 48:4, 903–934 http:www.aeaweb.org/articles.php?doi=10.1257/jel.48.4.903 903 1. Introduction G lobal warming is one of the most criti- cal, and also most daunting, challenges facing policymakers in the twenty-first cen- tury, (e.g., World Bank 2010). Assessing a globally efficient time path for pricing or controlling greenhouse gas (GHG) emis- sions is difficult enough, with huge scientific uncertainties, disagreement over the ulti- mate goals of climate policy, and disagree- ment over which countries should bear most responsibility for emissions reduc- tions. On top of this, domestic policy design is inherently difficult because of multiple, and sometimes conflicting, criteria for pol- icy evaluation. And at an international level, there are multiple approaches to coordinat- ing emissions control agreements. What should be a rational policy response for such an enormously complex problem? This paper attempts to provide some broad answers to this question, and to pin- point the main sources of controversy, by pulling together key findings from diverse literatures on mitigation costs, damage valuation, policy instrument choice, tech- nological innovation, and international cli- mate policy. Given that our target audience is the broader economics profession (rather than the climate specialist), our discussion is highly succinct and avoids details. We begin with the broadest issue of how much action to price or to control GHGs Designing Climate Mitigation Policy Joseph E. Aldy, Alan J. Krupnick, Richard G. Newell, Ian W. H. Parry, and William A. Pizer * This paper provides (for the nonspecialist) a highly streamlined discussion of the main issues, and controversies, in the design of climate mitigation policy. The first part of the paper discusses how much action to reduce greenhouse gas emissions at the global level is efficient under both the cost-effectiveness and welfare-maximizing paradigms. We then discuss various issues in the implementation of domestic emis- sions control policy, instrument choice, and incentives for technological innovation. Finally, we discuss alternative policy architectures at the international level. (JEL Q54, Q58) * Aldy: Special Assistant to the President on Energy and the Environment; his work on this paper was com- pleted while he was a full-time fellow at Resources for the Future. Krupnick: Resources for the Future. Newell: Nicholas School of the Environment and Earth Sciences, Duke University. Parry: Resources for the Future. Pizer: his work on this paper was completed while he was a full-time senior fellow at Resources for the Future. The authors are grateful to Carolyn Fischer, Roger Gordon, Charles Kolstad, Knut Rosendahl, Kenneth Small, and Brent Sohngen for helpful comments and to Michael Eber for research assistance.
35
Embed
Designing Climate Mitigation Policy - Harvard University
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Designing Climate Mitigation Policy903
1. Introduction
Global warming is one of the most criti- cal, and also most
daunting, challenges
facing policymakers in the twenty-first cen- tury, (e.g., World
Bank 2010). Assessing a globally efficient time path for pricing or
controlling greenhouse gas (GHG) emis- sions is difficult enough,
with huge scientific uncertainties, disagreement over the ulti-
mate goals of climate policy, and disagree- ment over which
countries should bear most responsibility for emissions reduc-
tions. On top of this, domestic policy design is inherently
difficult because of multiple, and sometimes conflicting, criteria
for pol- icy evaluation. And at an international level,
there are multiple approaches to coordinat- ing emissions control
agreements. What should be a rational policy response for such an
enormously complex problem?
This paper attempts to provide some broad answers to this question,
and to pin- point the main sources of controversy, by pulling
together key findings from diverse literatures on mitigation costs,
damage valuation, policy instrument choice, tech- nological
innovation, and international cli- mate policy. Given that our
target audience is the broader economics profession (rather than
the climate specialist), our discussion is highly succinct and
avoids details.
We begin with the broadest issue of how much action to price or to
control GHGs
Designing Climate Mitigation Policy
Joseph E. Aldy, Alan J. Krupnick, Richard G. Newell, Ian W. H.
Parry, and William A. Pizer*
This paper provides (for the nonspecialist) a highly streamlined
discussion of the main issues, and controversies, in the design of
climate mitigation policy. The first part of the paper discusses
how much action to reduce greenhouse gas emissions at the global
level is efficient under both the cost-effectiveness and
welfare-maximizing paradigms. We then discuss various issues in the
implementation of domestic emis- sions control policy, instrument
choice, and incentives for technological innovation. Finally, we
discuss alternative policy architectures at the international
level. (JEL Q54, Q58)
* Aldy: Special Assistant to the President on Energy and the
Environment; his work on this paper was com- pleted while he was a
full-time fellow at Resources for the Future. Krupnick: Resources
for the Future. Newell: Nicholas School of the Environment and
Earth Sciences, Duke University. Parry: Resources for the Future.
Pizer:
his work on this paper was completed while he was a full-time
senior fellow at Resources for the Future. The authors are grateful
to Carolyn Fischer, Roger Gordon, Charles Kolstad, Knut Rosendahl,
Kenneth Small, and Brent Sohngen for helpful comments and to
Michael Eber for research assistance.
Journal of Economic Literature, Vol. XLVIII (December
2010)904
is warranted in the near and longer term at a global level. There
are two distinct approaches to this question. The cost-effec-
tiveness approach acknowledges that policy- makers typically have
some ultimate target for limiting the amount of projected climate
change or atmospheric GHG accumulations, and the question is what
policy trajectory might achieve alternative goals at minimum
economic cost, accounting for practical con- straints, such as
incomplete international coor- dination. The other approach is to
weigh the benefits and costs of slowing climate change, which
introduces highly contentious issues in damage valuation, dealing
with extreme cli- mate risks, and intergenerational
discounting.
The second part of the paper deals with issues in the
implementation of climate pol- icy. At a domestic (U.S.) level,
these include a comparison of alternative emissions con- trol
instruments and how they should be designed to simultaneously
promote admin- istrative ease and minimize efficiency costs in the
presence of other policy distortions, abatement cost uncertainty,
and possible distributional constraints. We also discuss the extent
to which additional policies are warranted to promote the
development and deployment of emissions-saving technolo- gies. And
we briefly summarize emerging literature on alternative
international policy architectures. A final section discusses key
areas for future research.
2. Policy Stringency
2.1. Emissions Pricing to Stabilize Global Climate
The cost-effectiveness approach to global climate policy uses
models of the economic and climate system (known as integrated
assessment models) to estimate the emis- sions price trajectory
that minimizes the discounted worldwide costs of emissions
abatement, subject to a climate stabilization
target and possibly other, practical con- straints like delayed
developing country participation. These models range from bot-
tom-up engineering–economic models with considerable detail on
adoption and use of energy technologies to computable general
equilibrium models with a more aggregated and continuous structure
that better repre- sents demand responses, capital dynamics, and
factor substitution. Many models are hybrids containing substantial
technological detail in the energy sectors and more aggre- gate
representation in others. Typically the suite of existing and
emerging technologies is taken as given, although some models
capture induced innovation through learn- ing-by-doing and a few
have incorporated R&D-based technological change (e.g.,
Lawrence H. Goulder and Koshy Mathai 2000).
The choice of model structure is gener- ally less important than
assumptions about future baseline data and technology options.
Future mitigation costs are highly sensitive to business-as-usual
(BAU) emissions, which depend on future population and GDP growth,
the energy-intensity of GDP, and the fuel mix. They also depend on
the future availability and cost of emissions-saving tech- nologies
like nuclear and renewable power, carbon capture and storage, and
alternative transportation fuels. Considerable uncer- tainty
surrounds all of these factors.
Given the difficulty of judging which mod- els give the most
reliable predictions, we discuss a representative sample of
results, beginning with studies that assume emissions reductions
are efficiently allocated across countries and time, and use the
least expen- sive technological options (this is known as “where,
when, and how” flexibility). The results, summarized in table 1,
are from the U.S. Climate Change Science Program (CCSP, Product
2.1A), based on results from three widely regarded models (see Leon
E. Clarke et al. 2007 for details), and from the
905Aldy, Krupnick, Newell, Parry, and Pizer: Designing Climate
Mitigation Policy
Stanford Energy Modeling Forum’s EMF- 21 study (reported in
Francisco C. de la Chesnaye and John P. Weyant 2006) based on
sixteen models.
2.1.1. Reference Scenarios
Global CO2 emissions from fossil fuels have grown from about 2
billion (metric) tons in 1900 to current levels of about 30 bil-
lion tons and, in the absence of mitigation policy, are projected
to roughly triple 2000 levels by the end of the century (table 1).
The huge bulk of the projected future emissions growth is in
“non-Annex 1” (nonindustrial) countries—CO2 emissions from these
coun- tries have just overtaken those from “Annex 1” (industrial)
countries.1 These rising emis- sions trends reflect growing energy
demand from population and real income growth outweighing energy-
and emissions-saving technological change—traditional fossil fuels
still account for around three-quarters of global primary energy
consumption by 2100 (Clarke et al. 2007, table TS1).2
About 55 percent of CO2 releases are immediately absorbed by the
upper oceans and terrestrial biosphere while the remain- der enters
the atmosphere and is removed by the ocean and terrestrial sinks
only very gradually (Intergovernmental Panel on Climate Change
2007). The longer term rate of removal of CO2 from the atmosphere
is around 1 percent a year (i.e., CO2 has an expected atmospheric
residence time of
1 The 1990 U.N. Framework Convention on Climate Change grouped
countries into either Annex 1 or non- Annex 1 according to their
per capita income at that time. Only Annex 1 countries agreed to
reduce emissions under the 1997 Kyoto Protocol.
2 Land-use changes currently contribute about an additional 5.5
billion tons of CO2 releases (primarily through deforestation in
developing countries for agri- culture and timber) though these
sources are projected to grow at a much slower pace than fossil
fuel emissions (Intergovernmental Panel on Climate Change 2007).
Land-use CO2 emissions are not priced in the models in table
1.
about a century), and even this very gradual decay rate might
decline as oceans become more saturated with CO2. Stabilizing atmo-
spheric CO2 concentrations over the very long term essentially
requires elimination of fossil fuel and other GHG emissions.
Atmospheric CO2 concentrations in creased from preindustrial levels
of about 280 parts per million (ppm) to 384 ppm in 2007, and are
projected to rise to around 700–900 ppm by 2100 (table 1).
Accounting for non-CO2 GHGs, such as methane and nitrous oxides
from agriculture, and expressing them on a lifetime warming
equivalent basis, the CO2-equivalent concentration is about 430 ppm
(Intergovernmental Panel on Climate Change 2007). Total GHG
concentrations in CO2-equivalents are projected to reach 550 ppm
(i.e., about double preindustrial levels) by around mid
century.
Globally averaged surface temperature is estimated to have risen by
0.74°C between 1906 and 2006, with most of this warming due to
rising atmospheric GHG concentra- tions, as opposed to other
factors like changes in solar radiation, volcanic activity, and
urban heat absorption (Intergovernmental Panel on Climate Change
2007). Figure 1, from Intergovernmental Panel on Climate Change
(2007), shows the projected long run warm- ing associated with
different stabilization levels for atmospheric CO2-equivalent con-
centrations (the climate system takes several decades to fully
adjust to changing concen- tration levels, due to gradual heat
diffusion processes in the oceans). If CO2-equivalent
concentrations were stabilized at 450, 550, and 650 ppm, mean
projected warming over pre-industrial levels is 2.1, 2.9, and 3.6°C
respectively. Figure 1 also indicates “likely ranges” of warming
about the mean projec- tion, which refer to an approximate 66 per-
cent confidence interval, based on sensitivity analysis from
scientific models—for example, the likely warming range for 550 ppm
CO2- equivalent stabilization is 1.9–4.4°C. The
Journal of Economic Literature, Vol. XLVIII (December
2010)906
TABLE 1 Least-Cost Policies to Stabilize Global Climate
2025 2050 2100
Mini- CAM IGSM MERGE
Mini- CAM IGSM
Global CO2 emissions, relative to 2000 Reference 1.27 1.46 1.70
1.59 1.98 2.59 3.42 3.21 3.45 450 CO2 stabilization 0.92 0.97 0.86
0.53 0.57 0.64 0.24 0.39 0.55 550 CO2 stabilization 1.25 1.35 1.22
1.32 1.56 1.20 0.79 0.71 0.81
CO2 concentration, ppmb
Reference 422 430 436 485 507 544 711 746 875 450 CO2 stabilization
412 416 408 434 440 430 426 456 451 550 CO2 stabilization 421 427
421 478 490 472 535 562 526
CO2 price, $/tonc
450 CO2 stabilization 41 36 88 157 127 230 166 173 1,651 550 CO2
stabilization 3 6 26 10 19 67 127 115 475
% reduction in world GDPd
450 CO2 stabilization 0.8 0.5 2.6 1.8 1.6 5.4 1.4 1.4 16.1 550 CO2
stabilization 0.0 0.0 0.7 0.2 0.2 1.8 0.7 1.0 6.8
U.S. CO2 emissions, relative to 2000 Reference 1.25 1.10 1.40 1.27
1.20 2.00 1.63 1.34 2.93 450 CO2 stabilization 0.79 0.83 0.88 0.42
0.43 0.54 0.02 0.27 0.40 550 CO2 stabilization 1.24 1.05 1.04 1.02
0.98 1.13 0.29 0.37 0.59
EMF-21e lower end median upper end lower end median upper end lower
end median upper end
Global CO2 emissions, relative to 2000 Reference 1.33 1.48 1.64
1.64 1.88 2.23 2.11 2.93 3.52 550 CO2 stabilization 1.17 1.25 1.41
1.13 1.25 1.41 0.66 0.90 1.25
CO2 price, $/tonc
550 CO2 stabilization 3 13 21 12 33 99 31 92 166
% reduction in world GDP d
550 CO2 stabilization 0.1 0.1 0.8 0.2 0.6 3.1 0.3 5.1 8.2
U.S. CO2 emissions, relative to 2000 Reference 1.19 1.26 1.38 1.31
1.65 1.97 0.95 1.85 2.29 550 CO2 stabilization 1.05 1.14 1.22 0.76
1.02 1.26 0.36 0.53 1.05
Notes: a Results are from the Integrated Global Systems Model
(IGSM), the Model for Evaluating Regional and Global Effects
(MERGE), and MiniCAM Model. See Clarke et al. (2007) for
details.
b The models stabilize concentrations of all GHGs, rather than CO2
alone (i.e., the CO2-equivalent concen- tration level is higher
than the CO2 concentration). Actual CO2 concentrations may
temporarily overshoot the long run targets.
c In year 2000 dollars or thereabouts. d GDP losses are not broken
out by region in the models. Losses include those from pricing CO2
and other
GHGs on an equivalent basis. The figures do not account for the
benefits of reduced climate change. e Modeling results from
Stanford’s Energy Modeling Forum, reported in de la Chesnaye and
Weyant (2006).
The results are from 16 models for CO2 prices and 12 models for
GDP. Lower and upper ends correspond to lower and upper two-thirds
of model results. Atmospheric CO2 concentrations are not
reported.
907Aldy, Krupnick, Newell, Parry, and Pizer: Designing Climate
Mitigation Policy
fundamental concern is that warming might greatly exceed these
ranges due to poorly understood feedbacks not represented in these
models, such as heat-induced releases of methane stored under the
oceans and in the permafrost.
2.1.2. Least-Cost Pricing
Most economic analysis has focused on climate stabilization targets
that are approxi- mately consistent with limiting atmospheric CO2
concentrations to either 450 or 550 ppm (with other GHGs included,
CO2-equivalent concentrations stabilized at approximately 530 and
670 ppm respectively). The studies
in table 1 examine globally cost-effective pricing of all GHGs that
are approximately consistent with these goals.3
Across the models and stabilization scenar- ios in table 1, CO2
emissions prices (in year 2000 dollars) rise steadily (beginning
around
3 The G-8 countries recently adopted a target of limit- ing
projected warming to 2oC above preindustrial levels. This would
require ultimately stabilizing CO2-equivalent concentrations at 450
ppm, which is considerably more stringent than the 450 ppm CO2
target discussed here. In fact, with current technologies, it is
difficult to see how the more stringent target could be achieved
(even allowing for transitory overshooting), given that current
concentration levels are already approaching this target.
E qu
ili br
iu m
te m
pe ra
tu re
in cr
ea se
o C
350 550 750 950 1150
– – – – –
Figure 1. Steady State Warming above Preindustrial Temperatures
from Stabilization at Different GHG Concentrations
Note: The black curve indicates the central case projection and the
grey curves indicate the 66 percent confidence interval.
Source: International Panel on Climate Change (2007), table
10.8.
Journal of Economic Literature, Vol. XLVIII (December
2010)908
year 2012) at approximately 5 percent a year, where this figure is
the consumer discount rate plus the atmospheric CO2 decay rate
(Stephan C. Peck and Y. Steve Wan 1996). However, one striking
feature in table 1 is the considerable price variation across mod-
els within a stabilization scenario, reflecting different
assumptions about future BAU emissions growth and future costs of
carbon- saving technologies. The other striking fea- ture is the
dramatic differences between the 550 and 450 ppm CO2 stabilization
targets. In the 550 ppm case, CO2 prices are $3–26 and $10–99 per
ton in 2025 and 2050 respec- tively, with global emissions 17–41
percent and 13–56 percent above 2000 levels at these dates,
respectively. In the 450 ppm case, CO2 prices are 3–16 times those
in the 550 ppm case to mid century, while emissions are 3–14
percent and 36–47 percent below 2000 levels in 2025 and 2050
respectively.4
Although GDP losses may be an unreliable proxy for efficiency
losses we discuss them here as they are the least common denomi-
nator reported by the modeling groups. Under the 550 ppm CO2
target, most mod- els project global GDP losses (from reducing both
CO2 and non-CO2 GHGs) of less than 1 percent out to 2050, though
some models suggest GDP losses could reach 2–3 percent by this
date. In present value terms, these losses amount to about $0.4–12
trillion out to 2050 when applied to a world GDP that is $60
trillion and growing (Richard G. Newell 2008, p. 12). Under the 450
ppm CO2 target, GDP losses are about 1.0–2.5 percent and 1.5–5.5
percent in 2025 and 2050 respec- tively or about $8–43 trillion in
present value from 2010 to 2050.
Under both 450 and 550 ppm CO2 sta- bilization scenarios, the
energy system is transformed over the next century (though
4 Some analysts express prices per ton of carbon rather than CO2.
To convert to $ per ton of carbon, multiply by the ratio of
molecular weights, 44/12=3.67.
at very different rates), through energy conservation, improved
energy efficiency, and particularly reductions in the carbon
intensity of energy. Most of the emissions reductions in the first
two to three decades occur in the power sector, largely through the
progressive replacement of traditional coal plants by coal with
carbon capture and storage, natural gas, nuclear, and renew- ables
(wind, solar, and biomass). However, the projected fuel mix is
highly sensitive to speculative assumptions about the relative
costs and availability of future technologies. For example, there
are considerable prac- tical obstacles to the expansion of nuclear
power (because of safety issues), renewables (because sites are
typically located far from population centers), and carbon capture
and storage (because of the difficulty of assigning sub-surface
property rights).5
As for U.S. CO2 emissions, in the BAU case they increase by about
30–100 percent above 2000 levels (of approximately 6 billion tons)
by mid century (table 1). Under the 550 CO2 ppm target, emissions
initially rise, then fall to roughly 2000 levels by 2050, and fall
rapidly thereafter. Under the 450 ppm target, U.S. emissions are
rapidly reduced to roughly half 2000 levels by 2050.6 U.S.-
specific GDP losses are not reported in the studies in table 1, but
allocating a quarter of the global cost to the United States (based
on its share in global GDP) implies a present
5 The transition away from coal reflects not only the range of
substitution possibilities in the power sector, but also the
disproportionately large impact of emissions pric- ing on coal
prices. A $10 price per ton of CO2 in the United States would
increase 2007 coal prices to utilities by about 60 percent,
wellhead natural gas prices by 9 percent, retail electricity and
crude oil prices each by 7 percent, and gaso- line prices by 3
percent (from Clarke et al. 2007, table TS5, and
www.eia.gov).
6 As of 2009, proposed climate policies in the United States embody
emission reduction targets approximately equivalent to about of 80
percent below 2000 levels by 2050. However, actual reductions in
U.S. CO2 emissions would be about 60 percent if provisions to use
domestic and international emission offsets were fully
exploited.
909Aldy, Krupnick, Newell, Parry, and Pizer: Designing Climate
Mitigation Policy
value cost to the United States through mid century of about $0.1–3
trillion (0–1 percent of the present value of GDP) for the 550 ppm
target and $2–11 trillion (1–3 percent of present value GDP) for
the 450 ppm target.7
2.1.3. Deviations from Least-Cost Pricing
Aside from the uncertainty surrounding modeling assumptions, a key
qualification to the studies in table 1 is that they assume
globally efficient abatement policies. More likely, particularly
given the “common but differentiated responsibilities” recognized
in the Kyoto Protocol, participation in global mitigation efforts
among major developing country emitters will be delayed, causing
marginal abatement costs to differ across regions. For a given
climate stabilization scenario, to what extent does this affect
worldwide abatement costs and appropriate policies in developed
countries?
James A. Edmonds et al. (2008) explore these issues assuming Annex
1 countries agree to impose a harmonized emissions price starting
in 2012, China joins the agree- ment at a later date, and other
countries join whenever their per capita income reaches that of
China at the time of China’s accession. In one scenario, they
assume new entrants immediately face the prevailing Annex 1
emissions price, while in another the emis- sions price for late
entrants converges gradu- ally over time to the Annex 1 price. The
analysis accounts for emissions leakage, that is, the increase in
emissions in nonparticipat- ing countries due to the global
relocation of energy-intensive firms, and increased use of
7 U.S.-specific models project emissions price ranges that are
broadly consistent with those in table 1. For example, analyses by
Sergey Paltsev et al. (2007), U.S. Environmental Protection Agency
(2008), U.S. Department of Energy, Energy Information
Administration (2008a), and CRA International (2008) project
emissions prices of around $40–90 per ton of CO2 in 2025 for
climate legisla- tion that would reduce U.S. CO2 emissions by about
20 percent below 2000 levels by that date.
fuels elsewhere as decreased demand in par- ticipating countries
lowers world fuel prices.
Under the 550 ppm CO2 target, even if China joins between 2020 and
2035, the implications for Annex 1 policies can be sig- nificant
but are not that striking. Compared with the globally efficient
policy, near-term Annex 1 emissions prices rise from between a few
percent to 100 percent under the dif- ferent scenarios, and
discounted global abatement costs are higher by 10–70 per- cent.
However, under the 450 ppm CO2 target, essentially all of the
foregone earlier reductions in non-Annex 1 countries must be offset
by additional early reduction in Annex 1 countries (rather than
more global abate- ment later in the century). This can imply
dramatically higher near-term Annex 1 emis- sions prices,
especially with longer delay and lower initial prices for late
entrants. Under these scenarios, discounted global abate- ment
costs are about 30–400 percent higher than under globally efficient
pricing, and near and medium term emissions prices can be an order
of magnitude larger with China’s accession delayed till 2035.
A further key point from Edmonds et al. (2008) is the potentially
large shift in the global incidence of abatement costs, under-
lying the disincentives for early developing country participation.
In the globally effi- cient policy, without any international
trans- fer payments, developing countries bear about 70 percent of
discounted abatement costs out to 2100, while they bear “only”
17–34 percent of global abatement costs when China’s accession
occurs in 2035 and new entrants face lower starting prices.
Finally, insofar as possible pricing non-CO2 GHGs is also
important. According to mod- eling results in de la Chesnaye and
Weyant (2006), GDP costs are 20–50 percent larger when only CO2, as
opposed to all, GHGs are priced, for the same overall limit on
atmo- spheric CO2-equivalent concentrations. This reflects
opportunities for large-scale,
Journal of Economic Literature, Vol. XLVIII (December
2010)910
low-cost options for non-CO2 abatement in the first half of this
century, though practi- cal difficulties in pricing other GHGs are
not factored into the models.
2.1.4. Summary
There is a large difference in the appro- priate starting prices
for GHG emissions, depending on whether the ultimate objec- tive is
to limit atmospheric CO2 concentra- tions to 450 or 550 ppm—targets
that are approximately consistent with keeping the eventual, mean
projected warming above preindustrial levels to 2.7 and 3.7oC
respec- tively (assuming non-CO2 GHGs are also priced). The 450 ppm
target implies emis- sions prices should reach around $40–90 per
ton of CO2 by 2025, while the 550 ppm target implies prices should
rise to $3–25 by that date. Securing early and widespread partici-
pation in an international emissions control regime can also be
critical for containing costs under the 450 ppm target, while under
the 550 ppm target there is greater scope for offsetting the effect
of delayed participa- tion through greater emissions reductions in
the latter half of the century. Given the con- siderable difference
in GDP losses at stake between the two targets ($8–43 trillion in
present value under cost-effective pricing out to 2050 compared
with $0.4–12 trillion), it is important to carefully assess what
start- ing prices might be justified by avoiding cli- mate change
damages.
2.2. Welfare-Maximizing Emissions Pricing
2.2.1. Marginal Damage Estimates
Estimates of the marginal damages from current emissions begin with
a point estimate of total contemporaneous damages from warming,
usually occurring around 2100. Total damage estimates from a number
of studies are roughly in the same ballpark for a given amount of
warming. According to rep- resentative estimates in figure 2,
damages are
in the range of about 1–2 percent of world GDP for a warming of
2.5oC above preindus- trial levels, though some estimates are close
to zero or even negative (the prospects for negative costs
diminishes with greater warm- ing). For warming of about 4.0oC,
damage estimates are typically in the order of 2–4 percent of world
GDP. However, similarities in aggregate impacts mask huge
inconsisten- cies across these studies, which reach strik- ingly
different conclusions about the size of market and nonmarket damage
categories and expected catastrophic risks.
Very few studies attempt to value the dam- ages from more extreme
warming scenarios, given so little is known about the physical
impacts of large temperature changes. Two exceptions are William D.
Nordhaus and Joseph Boyer (2000) and Nicholas Stern (2007) who put
expected total damages at 10.2 and 11.3 percent of world GDP, for
warming of 6.0oC and 7.4oC respectively, though these figures are
necessarily based on extrapolations and subjective judgment. Again,
there is little consistency across the estimates. In Nordhaus and
Boyer (2000), catastrophic risks and market damages account for
about 60 and 40 percent of total damages respectively, with
nonmarket impacts roughly washing out (for example, the gains from
leisure activities offset losses from the disruption of ecosystems
and set- tlements). In contrast, nonmarket impacts account for
about half of Stern’s overall dam- age estimate.
Marginal damage estimates are based on assumptions about
emissions/concentra- tion relationships, climate adjustment and
sensitivity, damages from climate change (inferred from a point
estimate of total dam- ages using functional form assumptions), and
discount rates. Richard S. J. Tol (2009) conducts several
meta-analyses of marginal damage estimates, reporting median esti-
mates of $4.1–20.2 per ton of CO2 (individ- ual studies are not
independent however, as
911Aldy, Krupnick, Newell, Parry, and Pizer: Designing Climate
Mitigation Policy
they often draw from the same sources and from each other).
Although individual esti- mates are highly divergent, most are on
the low side (see also Stephen C. Newbold et al. 2009). Especially
striking is the difference between Stern (2007) at $85 and Nordhaus
(2008) at $8 per ton of CO2—a difference largely dependent on
discount rate assump- tions (see below).8
There is some consensus that marginal damages grow at around 2–3
percent a year in real terms (approximately the rate of growth in
output potentially affected by cli- mate change) or about half the
rate as under cost-effective emissions pricing. Marginal
8 Some of the differences in marginal damage esti- mates reflect
different assumptions about the year for which emissions are being
priced, and about the extent of future warming. Most estimates of
near-term Pigouvian taxes (i.e., marginal damages from the globally
optimized
damages rise with the extent of warming (suggesting a faster rate
of increase), but an offsetting factor is that warming is a concave
(logarithmic) function of atmospheric con- centrations. Although
CO2 concentrations ultimately reach 650 ppm in the twenty- second
century in Nordhaus’s (2008) optimal policy, constraining CO2
concentrations to 550 ppm affects, only modestly, the emis- sion
price trajectory to 2050. Thus, optimal near and medium term
emissions prices in Nordhaus (2008) are in the same ballpark with
those for cost-effective stabilization of CO2 concentrations at 550
ppm, while starting prices in Stern (2007) are broadly
emissions trajectory) are similar to marginal damage esti- mates at
BAU emissions levels. One exception is Stern (2007, p. 344) where
marginal damages are considerably reduced when aggressive climate
stabilization goals are achieved.
5
4
3
2
1
0
–1
–2
–3
Mendelsohn & Williams 2007*
Nordhaus 2008** Nordhaus 2008** Stern 2007** Tol 1995 Tol
2002**
Market Catastrophic Non-market
Figure 2. Selected Estimates of Contemporaneous World GDP Damages
from Global Warming Occurring around 2100
Notes: * Only market damages were estimated in these studies. The
above figure is the midpoint of a range of damage estimates. **
Market/nonmarket impacts are not precisely delineated in these
studies.
Journal of Economic Literature, Vol. XLVIII (December
2010)912
consistent with cost-effective prices to sta- bilize CO2
concentrations at 450 ppm, or lower.
2.2.2 Controversies in Marginal Damage Assessment
Differences in marginal damage estimates are largely explained by
fundamentally dif- ferent approaches to discounting rather than
differences in total damages from a given amount of warming
(Nordhaus 2007). However, the valuation of catastrophic and
noncatastrophic damages is also highly contentious.
Discounting. The descriptive approach to discounting argues that we
can do no better than using observed market rates, typically
assumed to be about 5 percent.9 According to this approach, market
rates reveal individuals’ preferences, as best we understand them,
about trade-offs between early and later consumption within their
lifecycle, as well as their ethical or intergen- erational
preferences. And they reflect the return earned by a broad range of
private and public investments—the opportunity cost against which
other, even intergenera- tional, investments ought to be measured.
Proponents of the descriptive approach view discounting at market
rates as essential for meaningful, consistent policy analysis and
to avoid highly perverse implications in other policy
contexts.
In contrast, the prescriptive approach argues that market rates
cannot be used when looking across cohorts (rather than
9 There are many market rates, from the long-term pre- tax real
return to equities (about 7 percent) to the after-tax return to
government bonds (about 2 percent). Converting all values into
their consumption equivalents, and dis- counting at the consumption
rate of interest, narrows the possible range of choice (e.g.,
Robert C. Lind 1982). In fact, Ellen R. McGrattan and Edward C.
Prescott (2003) suggest that the divergence in effective rates of
return is actually small, with an average real debt return during
peacetime over the last century of almost 4 percent and the average
equity return somewhat under 5 percent.
within individuals’ lifetimes). Instead, the discount rate (r) is
decomposed as follows: r = ρ + xη, where ρ is the pure rate of time
preference, x is the growth rate in consump- tion, and η is the
elasticity of marginal util- ity with respect to consumption. In
Stern (2007), for example, ρ = 0.1, x = 1.3, and η = 1, implying r
= 1.4. Choosing a value for ρ, the rate at which the utility of
future generations is discounted just because they are in the
future, is viewed as a strictly ethi- cal judgment. And ethical
neutrality, in this approach, essentially requires setting the pure
rate of time preference equal to zero. Discriminating against
people just because they are in the future is viewed as being akin
to discriminating against people in the present generation just
because they live in different countries (Geoffrey Heal 2009).
There is also controversy over the appropri- ate value for η, which
is almost as important as ρ. For example, Partha Dasgupta (2007)
argues for using a value of 2 to 4 on norma- tive grounds, while
Anthony B. Atkinson and Andrea Brandolini (2010) suggest a value
below unity is plausible, based on observed government
behavior.10
Catastrophic Risks. Although Nordhaus and Boyer (2000) and Stern
(2007) include catastrophic risks in their damage assess- ments,
the numbers are best viewed as highly speculative placeholders.
Nordhaus and Boyer (2000) put the annual willing- ness to pay to
avoid catastrophic risks at 1.0 and 6.9 percent of world GDP, for
warming levels of 2.5 and 6.0oC respectively, based on subjective
probabilities (from an expert
10 Besides ethical arguments, Thomas Sterner and U. Martin Persson
(2008) argue for discounting the non- market impacts of climate
change (e.g., ecosystem loss) at below market rates. This is
because the value of non- market goods (which are essentially fixed
in supply) rises over time relative to the value of market goods
(for which supply increases along with demand), assuming market and
nonmarket goods are imperfect substitutes for one another.
913Aldy, Krupnick, Newell, Parry, and Pizer: Designing Climate
Mitigation Policy
elicitation survey) for these warming levels permanently wiping out
about a third of world GDP. In his central case, Stern (2007)
assumes the chance of catastrophic climate change is zero up to a
warming of about 5oC, beyond which the annualized risk of regional
GDP losses of 5–20 percent rises by about 10 percent for each
additional 1oC of warming.
Martin L. Weitzman (2009a) takes a radi- cally different
perspective. He shows that, if the probability of increasingly
catastrophic outcomes falls more slowly than marginal utility in
those outcomes rises (with dimin- ished consumption), then the
certainty- equivalent marginal damage from current emissions
becomes infinite. These condi- tions apply if the probability
distribution for climate sensitivity is a fat-tailed t-distribution
(i.e., approaches zero at a less than exponen- tial rate) and
utility is a power function of consumption. Although marginal
utility is probably not unbounded, Weitzman shows that with
probabilities of a 20oC temperature change inferred from
Intergovernmental Panel on Climate Change (2007), and assuming this
temperature change would lower world consumption to 1 percent of
its current level, expected catastrophic damages could easily dwarf
noncatastrophic damages (even with these impacts delayed a century
or more and discounted at market rates).11
There are several responses to the Weitzman critique. One is that,
most likely, the probability distribution for climate sen- sitivity
may have thin rather than fat tails. If the distribution is
thin-tailed, Newbold
11 The Intergovernmental Panel on Climate Change report provides
probability distributions from twenty- two scientific studies.
Combining these distributions, Weitzman (2009a) suggests that there
is a 5 percent and 1 percent probability that eventual warming from
a doubling of CO2 equivalent concentrations will exceed 4.5°C and
7.0°C respectively. However, making an (extremely crude) adjustment
for the possibility of feedback effects he infers a distribution
where the probability of eventual tempera- ture change exceeding
10°C and 20°C is 5 percent and 1 percent respectively.
and Adam Daigneault (2009) and Robert S. Pindyck (2008) find that
damage risks from extreme global warming are typically under 3
percent of consumption (rather than infi- nitely large).
Second, setting a modest emissions price now does not preclude the
possibility of a mid-course correction, involving a rapid
phase-down in global emissions, should future learning reveal we
are on a cata- strophic trajectory (e.g., Gary W. Yohe and Tol
2009). This argument assumes policy- makers can avoid the
catastrophe—it breaks down if this would require reversing previ-
ous atmospheric accumulations because an abrupt climate threshold
has been crossed.
Finally, a costly, rapid stabilization of GHG concentrations is a
highly inefficient way to address the very small probabil- ity of
extreme outcomes, if a portfolio of last-resort technologies could
be success- fully developed and deployed, if needed, to head off
the catastrophe. These include “air capture” technologies for
atmospheric GHG removal and “geo-engineering” tech- nologies for
modifying global climate.12 Moreover, these R&D efforts can be
led by one or several countries, avoiding the challenges endemic in
organizing a rapid emissions phasedown among a large num- ber of
emitting countries with widely dif- fering interests. Nonetheless,
public R&D into last-resort technologies (virtually non-
existent at present) is highly contentious. One objection is that
advancing last-resort technologies could undermine support for
emissions mitigation efforts. Another is that geo-engineering
(though not air capture)
12 Besides rapid reforestation programs, air capture might involve
bringing air into contact with a sorbent material that binds
chemically with CO2 and extraction of the CO2 from the sorbent for
underground, or other, dis- posal. Geo-engineering technologies
include, for example, deflection of incoming solar radiation
through shooting particles into the stratosphere or blowing oceanic
water vapor to increase the cover of reflective clouds.
Journal of Economic Literature, Vol. XLVIII (December
2010)914
could have extreme downside risks (e.g., from overcooling the
planet or radically altering precipitation patterns) that may be
difficult to evaluate prior to widespread deployment. Whether
effective institutions could be developed to prevent unilateral
deployment of climate modification tech- nologies prior to rigorous
assessment of their risks is also unclear (e.g., Scott Barrett
2008; David G. Victor 2008).
In short, the implications of extreme cata- strophic risks for
emissions pricing are highly controversial. So long as there is
some posi- tive likelihood, no matter how small, that the climate
sensitivity function is fat-tailed then catastrophic risks can
still swamp non- catastrophic impacts. Mid-course policy
corrections may come too late to prevent a catastrophe, given that
it may take several decades for the full warming impacts of pre-
vious atmospheric accumulations to be real- ized. And the future
viability of last-resort technologies is highly uncertain at
present. All of these issues—the nature and extent of damages from
extreme warming, the feasibil- ity of future, mid-course policy
corrections, and the efficient balance between mitigation and
investment in last-resort technologies— are badly in need of
economic analysis.
Noncatastrophic Impacts. Although on a different scale than
catastrophic risks, controversies abound in the valuation of
noncatastrophic damages. These include agricultural impacts, costs
of increased storm intensity and protecting against rising sea
levels, health impacts from heatwaves and the possible spread of
vector-borne disease, loss of ecosystems, and so on. Box 1 pro-
vides a very brief summary of attempts to value these damage
categories (see Michael Eber and Alan J. Krupnick 2009 for a more
detailed discussion). However, due to the rapid outdating of prior
research, daunting methodological challenges, and the small number
of economists working on aggregate damage assessment, the valuation
literature
remains highly inconsistent and poorly developed, as a few examples
illustrate (W. Michael Hanemann 2008).
Damage assessments (like those in figure 2) assume losses in
consumer and producer surplus in agricultural markets are equiva-
lent to anything from a net gain of about 0.1 percent to a net loss
of 0.2 of world GDP for warming of about 2.5oC occurring in 2100.
However more recent, country-specific evi- dence suggests that
output losses could be a lot larger than those assumed in the
damage assessments to infer welfare costs to agricul- ture. For
example, William R. Cline (2007) suggests total losses of
agricultural output in developing countries in the order of 30 per-
cent, while Raymond Guiteras (2008) esti- mates agricultural losses
of 30–40 percent for India. Even for the United States, Wolfram
Schlenker, Hanemann, and Anthony C. Fisher (2005) suggest that the
output of individual crops could fall by up to 70 per- cent by
2100. Similarly, recent evidence on ice melting suggests that sea
level rises over the next century may be more extreme than the
25–60 cm assumed in most previous damage assessments (box 1). And
estimated ecosystem losses of about 0.1–0.2 percent of world GDP
seem inconsistent with Andreas Fischlin et al.’s (2007) projection
that 20–30 percent of the world’s species (an enor- mous amount of
natural capital) faces some (though possibly slight) extinction
risk.
More generally, scientific models cannot reliably predict local
changes in average tem- perature, temperature variability, and
precip- itation, all of which are critical to crop yields. The
baseline for impact assessment decades from now is highly sensitive
to assumptions about regional development (including the ability to
adapt to climate change), future technological change (e.g., into
climate- and flood-resistant crops), and other policies (e.g.,
attempts to eradicate malaria or inte- grate global food markets).
Controversies surround the valuing of nonmarket effects
915Aldy, Krupnick, Newell, Parry, and Pizer: Designing Climate
Mitigation Policy
Box 1. Valuation of Noncatastrophic Climate Damages (for Warming of
2.5oC or Thereabouts Occurring Around 2100)
Agriculture. Estimates of consumer and producer surplus losses in
agricultural markets from pre- dicted changes in regional
temperature and precipitation use evidence on crop/climate
sensitivity from laboratory experiments and on regressions of land
values or farm performance on climate variables (e.g., Adams et al.
1990; Reilly et al. 2001; Mendelsohn et al. 1994, 2001). Laboratory
studies can control for confounding factors like soil quality and
the fertilizing effect of higher CO2 concentrations, while
regression analyses account for farm level adaptation (e.g.,
changes in crop variety and planting/harvesting dates). Worldwide
agricultural impacts have been built up using extrapolations from
U.S. studies, adjusting for differences in local agricultural
composition and climate, and, more recently, country-specific
evidence that captures local factors like adaptive ca- pability.
Studies show a pattern of gains in high latitude and temperate
regions (like Russia), where current temperatures are below optimum
levels for crop growth, counteracting damages in tropical regions,
where current temperatures are already higher than optimal.
Sea Level. The annualized costs of future global sea level rises,
due to thermal expansion and melt- ing of sea ice, have been
estimated using projections of which coastal regions will be
protected, engineering data on the costs of dikes, sea walls, beach
replenishment, etc., and estimated losses from abandoned or
degraded property in unprotected areas. Some studies assume
efficient behav- ior by local policymakers in their choice of which
areas to protect and at what time, while others assume all
currently developed areas will be protected (Yohe 2000). Nordhaus
(2008) also includes an estimate of property losses from increased
storm intensity due to greater wind speed and waves coming off a
higher water level. Whether storm frequency will increase with more
humid air is un- certain (IPCC 2007). Worldwide sea level impacts
have been extrapolated from U.S. evidence, ad- justing for the
fraction of local land area in close proximity to the coast, though
recently there have been some local studies that account for the
slope and elevation of coastal land and prospective population
growth (e.g., Ng and Mendelsohn 2005 on Singapore). Overall,
estimates are relatively modest, for example they amount to 0.32 of
world GDP in Nordhaus (2008).
Some scientists project that sea levels could increase by several
meters by 2100 (Hansen 2007) rather than the 25–60 cm projected by
IPCC (2007). This would have major impacts on New York, Boston,
Miami, London, Tokyo, Bangladesh, the whole of the Netherlands, and
so on, and would completely inundate several small island states.
Based on extrapolations from sea level protection costs in Holland,
the global costs of this more extreme sea level rise may be at
least an order of magnitude or more greater than for a moderate sea
level rise, especially if coastal protection can- not be
constructed expeditiously (Nicholls et al. 2008; Olsthoorn et al.
2008). Another possibility is that warming may cause changes in
ocean circulation patterns. However, IPCC (2007) projects that
warming from climate change will dominate any cooling effect on
Europe from a weaker Gulf Stream.
Other market sectors. Studies suggest other market impacts are
relatively minor. With most forests along the increasing part of
the inverted-U relation between forest productivity and
temperature, Sohngen et al. (2001) find positive overall impacts
from warming on global timber markets. Most studies find a net loss
for the energy sector, as increased costs for space cooling
dominate savings in space heating (e.g., Mendelsohn and Neumann
1999). Impacts on water availability also tend to be negative, as
increased evaporation reduces freshwater supplies, and the value of
these losses is compounded with greater demand for irrigation
(Mendelsohn and Williams 2007).
(continued)
Journal of Economic Literature, Vol. XLVIII (December
2010)916
(e.g., the value of mortality in poor coun- tries, how much people
in wealthy countries value ecosystem preservation in poor coun-
tries). There is scant evidence on additional risks, such as
extreme local climate change (e.g., from shifting monsoons and
deserts) and broader health effects (e.g., malnutri- tion from food
shortages, the net effects of milder winters and hotter summers,
and diarrhea if droughts reduce safe drinking water supplies). Most
of the impact assess- ment literature is based on extrapolations
from U.S. studies—country-specific studies that account for local
factors (e.g., ability to adapt farm practices to changing climate)
have only recently begun to emerge. Finally, worldwide results mask
huge disparities in
regional burdens, and there is disagreement on how to aggregate
impacts across regions with very different per capita
income.13
2.2.3 Further Issues Posed by Uncertainty
Finally, we touch on some additional com- plications for emissions
pricing posed by uncertain discount rates, risk aversion, and
irreversibility.
In damage valuation, the time path of future discount rates is
usually taken as
13 Most studies aggregate regional impacts using weights equal to
the region’s share in world GDP or world population. More
generally, use of distributional weights can increase total damage
estimates up to about 300 per- cent (e.g., David Pearce
2005).
Box 1 Valuation of Noncatastrophic Climate Damages (for Warming of
2.5oC or Thereabouts Occurring Around 2100) (continued)
Health. There have been some attempts to quantify future health
damages. For example, using sta- tistical evidence on climate and
disease, Nordhaus and Boyer (2000) put health risks from the possi-
ble spread of vector-borne diseases like malaria at 0.10 percent of
world GDP. Broader health risks are even more speculative.
According to McMichael et al. (2004), there were 166,000 excess
deaths worldwide in 2000 from climate change to date. Of these,
“only” 16 percent were from malaria, 46 percent reflected greater
malnutrition due to food shortages, another 28 percent more
diarrhea cases as droughts reduce safe drinking water supplies and
concentrate contaminants, while 7 per- cent were from temperature
extremes (most in Southeast Asia). However, malnutrition
projections are extremely sensitive to assumptions about whether,
over the next century, currently vulnerable regions develop, become
more integrated into global food markets, and are able to adopt
hardier crops. And increased incidence of water-borne illness might
be counteracted by future develop- ment and adoption of water
purification systems. Monetizing mortality effects is also
contentious as there are very few direct estimates of the value of
a statistical life for poor countries.
Ecosystems. All aspects of future climate change are potential
stressors to natural systems. Com- bining projections of ecosystems
at risk from climate change with evidence on the medicinal value of
plants and willingness to pay for species and habitat preservation,
Fankhauser (1995) and Tol (1995) put the value of ecosystem loss in
2100 at 0.21 and 0.13 percent of world GDP respectively. Nordhaus
and Boyer (2000) put the combined risks to natural ecosystems and
climate-sensitive human settlements at 0.17 percent of world GDP in
2100, assuming the capital value of vulnerable systems is 5−25
percent of regional output, and an annual willingness to pay equal
to 1 percent of capital value. These estimates are highly
speculative, given that very little is known about ecological
impacts and how people value large scale (as opposed to marginal)
ecosystem loss.
917Aldy, Krupnick, Newell, Parry, and Pizer: Designing Climate
Mitigation Policy
given. However, the discount factor applied to damages is a convex
function of the future discount rate, so discount rate uncer-
tainty (for a given expected value) increases the
certainty-equivalent discount factor (Weitzman 1998). Newell and
William A. Pizer (2003) estimated that discount rate uncertainty
(inferred from U.S. historical evidence) almost doubles estimates
of mar- ginal emissions damages.
Leaving aside extreme risks, should mar- ginal damage estimates
include a risk pre- mium? This would be appropriate if the marginal
utility of consumption, net of cli- mate damages, were larger in
high-damage outcomes, in which case a mean-preserving increase in
the spread of possible damages outcomes would increase expected
disutil- ity. However, if gross consumption is greater in
high-damage scenarios (for example, because rapid productivity
growth leads to both high consumption and high emission rates),
then the marginal utility of consump- tion net of damages is lower,
and possibly even lower than marginal utility in low- damage
states. Simulations by Nordhaus (2008, chapter 7) suggest this
might in fact be the case, implying the risk premium is actually
negative, though empirically small. On the other hand, we do not
know what the probability distribution over damage outcomes is. If
policymakers are averse to such ambiguity this may, under certain
con- ditions, imply a higher near term price on emissions, though
how much higher is diffi- cult to quantify (Andreas Lange and
Nicolas Treich 2008).
Returning to the issue of irreversibility and future learning, is
there an option value (which should be reflected in the emis- sions
price) gained from delaying atmo- spheric GHG accumulations until
more is known about how much damage they will cause? Option values
arise if such delay increases the potential future welfare gains
from responding to new information about
damage risk (Pindyck 2007). If damages are linear in GHG
concentrations, changes in the inherited concentration level do not
affect marginal damages from additional, future accumulations. In
this case, the wel- fare effects of policy interventions at differ-
ent time periods are decoupled (at least from the damage side), and
there is no option value. If instead, damages are convex in
atmospheric GHG accumulations the pros- pect of future learning
reduces the optimal near-term abatement level, to the extent that
the damages from near-term emissions can be lowered through greater
abatement in future, high-damage scenarios. Moreover, to the extent
that current abatement involves (nonrecoverable) sunk investments
in emis- sions-saving technologies, there is another source of
option value, from delaying long- lived emissions-saving
investments until more is known about the benefits of emis- sions
reductions (Charles D. Kolstad 1996a). For these reasons,
theoretical analyses sug- gest that the prospect of future learning
justifies less near-term abatement (Kolstad 1996b; Fisher and
Urvashi Narain 2003; Pindyck 2007). However, as already noted, the
critical exception to this is when there is a possibility of
crossing a catastrophic thresh- old in atmospheric concentrations
prior to future learning, which is essentially nonre- versible
given the nonnegativity constraint on future emissions.
2.2.4 Summary
Most estimates of near-term marginal damages are in the order of
$5–$25 per ton of CO2. This range is in the same ballpark as
near-term emissions prices consistent with least-cost stabilization
of atmospheric CO2 concentrations at 550 ppm. These prices
represent a lower bound on appropriate policy stringency. Much
higher prices (that are consistent with 450 ppm, or even more
stringent, CO2 stabilization targets) can be implied by low
discount rates and, possibly,
Journal of Economic Literature, Vol. XLVIII (December
2010)918
extreme catastrophic risks (depending on the shape of the climate
sensitivity distribution). Thus, whether moderate or aggressive
emis- sions pricing is currently warranted largely hinges on one’s
view of discounting, whether radical mid-course corrections in
response to future learning about catastrophes are fea- sible, and
the prospects for development of last-resort technologies.
3. Policy Design
3.1. Choice Among, and Design of, Domestic Emissions Control
Instruments
Debate over the choice of instrument for a nationwide carbon
control program is no longer about the superiority of market-based
approaches over traditional forms of regula- tion (like technology
mandates) but rather between the two market-based alternatives,
emissions taxes and cap-and-trade systems.14 In a world where the
emissions external- ity is the only market distortion, and there is
no uncertainty, either instrument could achieve the first-best
outcome, if the emis- sions cap at each date equals the emissions
that would result under the Pigouvian tax. Whether allowances are
auctioned or given away for free has distributional consequences
but does not affect efficiency in this setting, so long as firm
behavior does not influence their future allowance allocations. If
firms were free to bank and borrow emissions allowances, the
policies would still be equiva-
lent, if the permit trading ratios across differ- ent time periods
were equivalent to the ratio of Pigouvian emissions taxes at those
dates (Catherine Kling and Jonathan Rubin 1997).
The equivalence between the two instru- ments potentially breaks
down in the pres- ence of preexisting tax distortions, when
distributional impacts are a concern, and when there is
uncertainty. Despite these complications, to a large extent permit
sys- tems can be designed to mimic the effect of a tax, and vice
versa, and therefore the choice of instrument per se is less
important than whether the chosen instrument is well designed
(Goulder 2009). Aside from policy stringency, key design features
relate to the point and scope of regulation, the allocation of
policy rents, and possible provisions to limit price
volatility.
3.1.1 Point of Regulation
Either a CO2 tax or cap-and-trade sys- tem can be imposed upstream
where fuels enter the economy (the minemouth for coal or wellhead
for oil and natural gas) accord- ing to a fuel’s carbon content or,
as in the European trading program, to downstream emitters at the
point where fuels are com- busted. Upstream systems would require
monitoring some 2,000–3,000 entities in the United States or
European Union, while downstream systems would apply to 10,000 or
more power plants and large indus- trial smokestacks (Daniel S.
Hall 2007).15
14 Market-based instruments equalize marginal abate- ment costs
across all abatement opportunities within the firm, across
heterogeneous firms, across production sec- tors, and across
households and firms, by establishing an economy-wide emissions
price (J. H. Dales 1968; Allen V. Kneese and Blair T. Bower 1968;
William J. Baumol and Wallace E. Oates 1971; W. David Montgomery
1972). In contrast, for example, a requirement that all electric
utili- ties generate a fraction of their power from renewables will
not achieve any of these efficiency conditions. Some opportunities
at the firm level (e.g., substituting natural gas and nuclear power
for coal), are not exploited; marginal costs will differ across
heterogeneous power companies;
household electricity prices will not reflect the cost of the
remaining (unpriced) emissions; and abatement opportu- nities
outside of the power sector are unexploited. For a broad reviews of
the literature on environmental policy instrument choice, see
Cameron Hepburn (2006) and Goulder and Ian W. H. Parry
(2008).
15 If introduced at the same points in the economy, CO2 taxes and
cap-and-trade systems are likely to have very similar
administrative costs. Under cap-and-trade, costs also include those
from administering trading markets, as well as the transactions
costs of the trades themselves, though these are relatively small
(Robert N. Stavins 1995).
919Aldy, Krupnick, Newell, Parry, and Pizer: Designing Climate
Mitigation Policy
For a given total emissions reduction, the estimated economic costs
of downstream programs out to 2030 are not dramati- cally larger
than those for comprehensive upstream systems—about 20 percent
larger according to Goulder (2009)—even though downstream programs
cover only about half of total U.S. and EU CO2 emissions. This is
because the huge bulk of low-cost abatement opportunities are
(initially) in the power sector. Moreover, the infeasibil- ity of
monitoring emissions from vehicles, home heating fuels, and
small-scale indus- trial boilers in a downstream system can be
largely addressed through supplementary midstream measures targeted
at refined transportation and heating fuels, which fur- ther
narrows the cost discrepancy between upstream and downstream
systems.
There are a couple of other notable dif- ferences between the two
systems. One is that upstream programs must be combined with a
crediting system to encourage devel- opment and adoption of carbon
capture and storage technologies at coal plants and industrial
sources. (The tax credit should equal the amount of carbon
sequestered, as measured by continuous emission monitor- ing
systems, times the emissions price). The other is that, at least
for the United States where many states retain cost-of-service
regulation, the opportunity cost of freely allocated emissions
allowances to electric utilities in a downstream system may not be
passed forward into higher generation prices. As a result,
incentives for electricity conservation could be a lot weaker,
result- ing in a significant loss of cost-effectiveness, compared
with upstream programs or down- stream programs with full allowance
auction- ing (Dallas Burtraw et al. 2001).
3.1.2 Scope of Regulation
Domestic programs that fail to cover embodied carbon in products
imported from countries with suboptimal or no emissions
controls may cause significant emissions leak- age. The problem is
most relevant for down- stream, energy-intensive firms competing in
global markets (e.g., chemicals and plastics, primary metals,
petroleum refining), where reduced production at home may be
largely offset by increased production in other coun- tries with
higher emissions intensity than in the United States. According to
some mod- els, as much as 15–25 percent of economy- wide U.S. CO2
reductions could be offset by extra emissions elsewhere, although
the majority of the leakage stems from changes in global fuel
prices rather than relocation of footloose capital (Sujata Gupta et
al. 2007; Mun S. Ho, Richard Morgenstern, and Jhih- Shyang Shih
2008; Carolyn Fischer and Alan K. Fox 2007, 2009). Possible policy
responses to the latter source of leakage include impos- ing taxes,
or permit requirements, accord- ing to embodied carbon in product
imports (and symmetrical rebates for exporters) or to subsidize the
output of leakage-prone indus- tries (e.g., through output-based
allocations of free emissions allowances). However, all these
approaches may run afoul of interna- tional trade
obligations.
Certain non-CO2 GHGs are easily moni- tored (e.g., vented methane
from under- ground coalmines, fluorinated gases used in
refrigerants and air conditioners) and could be directly integrated
into a CO2 mitigation program through taxes, or permit trading
ratios, reflecting their relative lifetime warm- ing potential.
Other gases are far more diffi- cult to monitor, and are better
incorporated, insofar as possible, through offset provisions, where
the onus falls on the individual entity to demonstrate valid
reductions relative to a credible baseline. For example, meth- ane
from landfills and livestock waste might be collected, using an
impermeable cover, and flared or used in onsite power genera- tion,
while nitrous oxide might be reduced through changes in tilling and
fertilizer use (e.g., Shih et al. 2006; Hall 2007).
Journal of Economic Literature, Vol. XLVIII (December
2010)920
Finally, CO2 abatement through forest car- bon sequestration (e.g.,
from reducing defor- estation, reforesting abandoned cropland and
harvested timberland, modifying harvest practices to reduce soil
disturbance) appears to be relatively cost effective. According to
Stavins and Kenneth R. Richards (2005), as much as 30 percent of
U.S. fossil fuel CO2 emissions might be sequestered at a cost of up
to about $20 per ton of CO2. Coupling a domestic mitigation program
with offset pro- visions for forest carbon sequestration will
require measuring regional forest inventories to establish
baselines, monitoring changes in forest use (through remote sensing
and ground-level sampling) relative to the base- line, and
inferring the emissions implications of these changes based on
sampling of local tree species and age. However, even if these
monitoring challenges can be overcome, fur- ther problems remain.
One is that, without an international program covering major for-
ested countries, domestic reductions can be offset through
emissions leakage via changes in world timber prices (Brian C.
Murray, Bruce A. McCarl, and Heng-Chi Lee 2002 estimate the
international leakage rate could be anywhere from less than 10
percent to over 90 percent depending on the type of activity and
location in the United States). Another is that sequestered carbon
in trees is not necessarily permanent if trees are later cut down,
decay or burn, requiring assign- ment of liability to either the
offset buyer or seller for the lost carbon.
3.1.3 Allocation of Policy Rents
In their traditional form, emissions taxes raise revenues for the
government, while cap-and-trade systems create rents for firms
receiving free allowance allocations. However, through allowance
auctions, cap- and-trade systems can generate comparable revenues
to a tax, while rents can be provided under a tax through
inframarginal exemp- tions for emissions or carbon content.
Under
either instrument, the fraction of policy rents accruing to the
government rather than pri- vate firms, and how revenues are used,
are extremely important for efficiency and dis- tributional
incidence.
Fiscal Linkages. The implications for emissions control policies of
preexisting tax distortions in factor markets have received
considerable attention in the broader envi- ronmental economics
literature (e.g., A. Lans Bovenberg and Goulder 2002), though these
distortions are typically not integrated into energy–climate
models. This raises two issues: to what extent is there a cost
saving from policies that raise revenues and use them to offset
distortionary taxes like income and payroll taxes, and to what
extent do mod- els that ignore prior tax distortions produce
inaccurate estimates of policy costs?
The efficiency gain from recycling rev- enues in other tax
reductions (relative to returning them lump sum or leaving policy
rents in the private sector) is simply the amount of revenue raised
times the marginal excess burden of taxation. Although there is
uncertainty over behavioral responses in factor markets, a typical
assumption is that the marginal excess burden of income taxes (with
revenue returned lump sum) is around $0.25 for the United States,
or per- haps as high as $0.40 if distortions in the pattern of
spending created by tax prefer- ences (e.g., for employer medical
insurance or homeownership) are taken into account. For modest
carbon policies, the efficiency gain from revenue recycling can be
large relative to the direct efficiency cost of the policy, or
Harberger triangle under the mar- ginal abatement cost schedule.
For example, if a $30 tax on U.S. CO2 emissions (currently about 6
billion tons) reduces annual emis- sions by 10 percent, the
Harberger triangle is $9 billion, while the revenue-recycling
benefit is roughly $40–65 billion per year.
However, this does not necessarily mean that revenue-neutral CO2
taxes, or auctioned
921Aldy, Krupnick, Newell, Parry, and Pizer: Designing Climate
Mitigation Policy
allowance systems, produce a “double divi- dend” by reducing the
costs of the broader tax system, in addition to slowing climate
change. There is a counteracting, “tax- interaction” effect (e.g.,
Goulder 1995). Specifically, the (policy-induced) increase in
energy prices drives up the general price level, which reduces real
factor returns, and thereby (slightly) reduces factor supply and
efficiency. Most analytical and numeri- cal analyses of
environmental tax shifts find that the tax-interaction effect
exceeds the revenue-recycling effect, implying no double dividend,
and that abatement costs are actually higher due to the presence of
preexisting tax distortions. A rough rule of thumb from these
models is that the costs of revenue-neutral emissions taxes are
about 15 percent greater, due to interactions with prior tax
distortions, implying the optimal tax is 15 percent lower than the
Pigouvian tax (e.g., Bovenberg and Goulder 2002). However, the cost
increase is far more sub- stantial for policies that do not exploit
the revenue recycling effect (i.e., cap-and-trade with free
allowance allocation or CO2 taxes with revenues not used to
increase economic efficiency). According to formulas derived in
Goulder et al. (1999), the increase exceeds 100 percent when the
emissions reduction is below 30 percent.16
More generally, there are many ways that carbon policy revenues
might be used, such as funding technology programs, cli- mate
adaptation projects, deficit reduction, energy efficiency programs,
rebates to elec- tricity consumers, and any number of com- plex
adjustments to the tax system, though the efficiency implications
of these recy- cling options are often not well understood.
Although in recent years there has been more interest in permit
auctions, in some cases it is unclear how the revenues will be
spent.17 Unless legislation accompanying car- bon policies
specifies offsetting reductions in other distortionary taxes, there
is ambiguity to what extent this shift implies a reduction in the
overall costs of carbon policies.
Distributional Considerations. The distri- butional impacts of
emissions control poli- cies are potentially important for both
equity and feasibility.
On equity grounds the difference between (revenue-neutral) CO2
taxes/auc- tioned allowances, and allowance systems with free
allocation to firms, can be quite striking. Under the latter
policy, permit rents are reflected in higher firm equity values,
and therefore (through dividend and capital gains income)
ultimately accrue to shareholders, who are concentrated in upper
income groups. Terry Dinan and
16 There are some caveats here. One is that the pro- portionate
increase in abatement costs may be much smaller in other countries
if tax wedges in factor mar- kets are smaller than those in the
United States, or if labor markets are dominated by institutional
wage set- ting (e.g., Francesco Bosello, Carlo Carraro, and Marzio
Galeotti 2001). Another is that the tax-interaction effect is
weaker if, due to regulated pricing and/or infra- marginal rents on
coal technologies that bear some of the burden of emissions
pricing, there is incomplete pass through of emissions prices into
electricity prices (Antonio M. Bento and Mark Jacobsen 2007; Parry
2005). Finally, the revenue-recycling effect can dominate the
tax-interaction effect when tax preferences cause sig- nificant
distortions or when a large share of revenues are used to cut taxes
on capital as opposed to labor
(see Parry and Bento 2000 and Bovenberg and Goulder 1997
respectively).
17 For example, in the first two phases of the European Union’s CO2
trading program (2005–07 and 2008–12), over 95 percent of the
allowances were given away free to exist- ing emissions sources.
However, partly in response to the large windfall profits earned by
power companies, the plan is to transition to full allowance
auctions for that sector by 2020, with the decision on how to use
revenues largely left to the member states (Jos Sijm, Karsten
Neuhoff, and Yihsu Chen 2006; Commission of the European
Communities 2008). In the Regional Greenhouse Gas Initiative in the
United States, covering power sector CO2 emissions from ten
Northeastern and Mid-Atlantic states, allowances are auctioned with
revenues earmarked for energy efficiency and other clean technology
programs.
Journal of Economic Literature, Vol. XLVIII (December
2010)922
Diane Lim Rogers (2002) estimated that, for a 15 percent reduction
in CO2 emis- sions, U.S. households in the lowest-income quintile
would be worse off on average by around $500 per year, while
households in the top-income quintile reap a net gain of around
$1,000 (i.e., increased stock- holder wealth overcompensates this
group for higher energy prices). This inequitable outcome could be
avoided under emis- sions taxes and auctioned allowance sys- tems
if revenues were recycled in income tax reductions tilted toward
the poor (e.g., Gilbert E. Metcalf 2009).
As regards feasibility, compensation for adversely affected
industries may be part of the political deal-making needed to first
initiate, and progressively tighten, emissions controls (e.g., A.
Denny Ellerman 2005). Compensation, through free allowance allo-
cation or tax relief, may be required for both formally regulated
sectors and downstream sectors vulnerable to higher energy prices
(e.g., energy-intensive firms competing in global markets).
However, given the ten- sion between providing industry compensa-
tion, and the fiscal and (household) equity reasons for raising
revenue, it is important to know how much compensation is needed to
keep firms whole. At least for a moder- ately scaled CO2 permit
system, only about 15–20 percent of allowances are needed to
compensate energy intensive industries for their loss of producer
surplus, so the huge bulk of the allowances could still be auc-
tioned (Bovenberg and Goulder 2001, Anne E. Smith, Martin T. Ross,
and Montgomery 2002). Although there are reasons for phas- ing out
compensation over time, firms may still be amenable to this if they
receive excess
compensation in the early years of the pro- gram (e.g., Stavins
2007).18
3.1.4 Price Volatility
Another reason CO2 taxes and cap-and- trade systems may produce
different out- comes stems from uncertainty over future abatement
costs reflecting, for example, uncertainty over energy prices,
technologi- cal advances, and substitutes for fossil fuels.
Price Versus Quantity Instruments in their Pure Form. If the goal
is welfare maximiza- tion, abatement cost uncertainty strongly
favors emissions taxes over cap-and-trade systems in their pure
form. This is most easily seen in a static setting where the
marginal benefits from abatement are con- stant. In this case, a
Pigouvian emissions tax automatically equates marginal benefits to
marginal abatement costs, regardless of the position of the
marginal abatement cost schedule. In contrast, when emissions are
capped to equate marginal benefits with expected marginal abatement
costs, ex post abatement will either be too high or too low
depending on whether the marginal abate- ment cost schedule is
higher or lower than expected (Weitzman 1974; Marc J. Roberts and
Michael Spence 1976; Yohe 1978).
This basic result carries over to a dynamic context with a sequence
of annual (Pigouvian) taxes or emissions caps, and where environ-
mental damages depend on the accumulated atmospheric stock of
emissions. Here, we have strong reasons to believe that the mar-
ginal benefits from global emissions reduc- tions are essentially
constant, as abatement in any one year has minimal impact on the
atmospheric stock. In fact, with abatement cost uncertainty,
simulation analyses suggest
18 One reason for phasing out allowance allocations is that they
must initially be based on a firm’s historical emis- sion rates
(prior to program implementation), which may be viewed as
increasingly unfair as firms grow or contract at dif- ferent rates,
or change their fuel mix, over time. However,
any updating of baselines based on firm performance will likely
introduce distortions in firm behavior (Knut Einar Rosendahl 2008).
Free allowance allocation may also retard the exit of inefficient
firms from an industry if firms lose their rights to future
allocations when they go out of business.
923Aldy, Krupnick, Newell, Parry, and Pizer: Designing Climate
Mitigation Policy
that discounted welfare gains under (globally imposed) CO2 taxes
might be several times those under (equivalently scaled) permits
(e.g., Pizer 2002; Michael Hoel and Larry Karp 2002). A
qualification to this is that the welfare advantage of taxes is
less pronounced if abatement cost shocks persist over time and the
emissions cap can be adjusted in response to those shocks (e.g.,
Karp and Jiangfeng Zhang 2005; Newell and Pizer 2003).
Stabilizing Allowance Prices. Emissions price volatility under
cap-and-trade systems can be contained by allowing firms to bank
permits when permit prices (and marginal abatement costs) are low,
and borrow per- mits from future periods when prevailing prices are
high. In fact, if banking and bor- rowing were completely unlimited
and cost- less, expected allowance prices would rise at the
interest rate, and the system would be largely equivalent to that
of an emissions tax growing at the interest rate. Alternatively,
through establishing appropriate ratios for trading permits across
time, the allowance price trajectory could mimic the growth in
marginal emissions damages over time (e.g., Kling and Rubin
1997).
In fact, most existing cap-and-trade sys- tems (e.g., the federal
SO2 and regional CO2 programs in the United States and the European
Union’s CO2 program) now incor- porate banking and borrowing
provisions, though in response to concerns about default risk,
borrowing is penalized through unfavor- able trading ratios and/or
quantitative limits. Harrison Fell, Ian A. MacKenzie, and Pizer
(2008) estimate that banking and borrowing provisions contained in
leading U.S. federal climate proposals obtain about one quarter to
one half of the cost savings from emissions taxes over equivalent
cap-and-trade systems without these provisions.
An alternative approach is to limit price volatility through a
“safety valve,” where the government sells additional permits at a
fixed price to prevent allowance prices
from rising above a ceiling price (e.g., Henry D. Jacoby and
Ellerman 2004). Expected welfare under this policy is maximized by
essentially designing it to mimic a Pigouvian tax—that is, setting
the safety valve price equal to marginal emissions damages and the
emissions cap tight enough so the safety valve binds nearly all the
time (Pizer 2002). Intermediate cases (with higher safety valve
prices and/or less stringent caps) generate intermediate welfare
gains between those of the pure tax and emissions quota. A fur-
ther alternative is a collar which combines a price ceiling with a
price floor. This approach encourages additional abatement when
allowance prices are low (to offset reduced abatement when
allowance prices are high) and avoids the potentially harmful
impacts of the price ceiling only on incentives to invest in
emissions-saving technologies. According to Fell, MacKenzie, and
Pizer (2008) the annualized cost savings between emissions taxes
and fixed emissions quotas in the United States would be about $4
billion for an emissions price of around $20 per ton of CO2, with
safety valves and price collars yielding intermediate cost
savings.
One final twist in instrument choice is that the price flexibility
afforded by a cap- and-trade system with (unhindered) allow- ance
borrowing and banking could actually be advantageous from a social
welfare perspective, when there is learning about future damages
and emissions taxes can only be adjusted at discrete intervals
(Murray, Newell, and Pizer 2009).19 Under the for- mer policy, new
information about damages will be immediately reflected in the time
path of current and expected future allow- ance prices, as
speculators anticipate an
19 Uncertainty over the marginal benefit schedule, in the absence
of learning, would not affect the choice between emissions taxes
and cap-and-trade because, on average, cumulated emissions
reductions, and hence expected envi- ronmental benefits, are the
same under both instruments (e.g., Stavins 1996).
Journal of Economic Literature, Vol. XLVIII (December
2010)924
adjustment of future emissions targets in response to that
information. In contrast, it may take some time before emissions
taxes can be adjusted to reflect new information, leaving emissions
prices suboptimal during the period of policy stickiness.
3.2 Promoting Technology Development and Diffusion
Several studies have demonstrated the central role that the
availability and cost of advanced energy technologies plays in
deter- mining the future costs of GHG emission targets (e.g.,
Clarke et al. 2006; Edmonds, Joseph M. Roop, and Michael J. Scott
2000; Kenneth Gillingham, Newell, and Pizer 2008). For example,
Clarke et al. (2006) found that if ambitious goals for technology
development are achieved, this can reduce discounted global
abatement costs by 50 per- cent or more. Establishing a price on
CO2 emissions is the single most important policy for encouraging
the innovation that might bring about advanced technology develop-
ment. However, additional measures to pro- mote applied R&D,
more basic research, and technology deployment, may be justified to
the extent they address market failures at different stages of the
innovation process.
3.2.1 R&D Policy
One market failure stems from the inabil- ity of private sector
inventors or innovators to fully appropriate spillover benefits to
other firms that might copy a new technology,
imitate around the technology if it is under patent, or otherwise
use knowledge about the technology to advance their own research
programs (Adam B. Jaffe, Newell, and Stavins 2003). Numerous
empirical stud- ies suggest that technology spillovers cause the
(marginal) social return to (commercial) R&D to be several
times the (marginal) pri- vate return.20
The appropriability problem implies that R&D incentives will be
suboptimal, even under Pigouvian emissions pricing. One response
would simply be to set emissions prices at a level higher than
warranted by externalities. However, this would generate efficiency
losses from excessive short-term abatement, and would not
differentiate incentives across technologies that might face very
different market impediments. In fact, no single instrument—either
emissions pricing or R&D incentives—can effectively correct
both the emissions externality and the knowledge appropriability
problem: using one instrument alone may involve considerably higher
costs than employing two complementary instruments (Fischer and
Newell 2008; Goulder and Stephen H. Schneider 1999).
Unfortunately, available literature pro- vides limited guidance on
the design of complementary R&D instruments. It is not clear
which instrument among, for instance, research subsidies,
strengthened patent rules, or technology prizes, is most effi-
cient, as this depends on the magnitude of
20 For example, Zvi Griliches (1992), Edwin Mansfield (1985),
Charles I. Jones and John C. Williams (1998). Although there is a
possibility of excessive competition for a given amount of
innovation rent, analogous to the excessive competition for
open-access resources, this problem is generally thought to be
dominated by the imperfect appropriability effect (Griliches 1992).
In fact, the problem of suboptimal innovation incentives may be
especially severe for GHG-saving technologies, compared with
commercial technologies. For example, skepticism over long-term
commitments to emissions pricing, and
the desirability of retaining policy discretion to respond to
future scientific knowledge, undermines the durable and substantial
incentives needed for encouraging GHG- saving technology
investments with high upfront costs. Limited patent lifetimes may
also discourage firms from launching R&D programs until a high
enough emissions price is established (Reyer Gerlagh, Snorre
Kvendokk, and Rosendahl 2008).
Still, efficiency gains from correcting the R&D market failure
appear to be smaller than those from correcting the CO2 emissions
externality (Parry, Pizer, and Fischer 2003).
925Aldy, Krupnick, Newell, Parry, and Pizer: Designing Climate
Mitigation Policy
technology spillovers, the scope for monop- oly pricing under
patents, and asymmetric information between governments and firms
about the expected benefits and costs of research (e.g., Brian
Davern Wright 1983). And just how much applied R&D in the
energy sector should be expanded is difficult to estimate, given
uncertainty over the pro- ductivity of research and the risk of
crowding out socially valuable research elsewhere in the economy
(e.g., Nordhaus 2002; Goulder and Schneider 1999).
3.2.2 Basic Research
Appropriability problems are most severe for more basic research,
which is largely conducted by universities, other nonprof- its, and
federal labs, mostly through central government funding. While it
is not practical to assess the efficient allocation of funding
across individual programs, Newell (2008, p. 32) suggests that a
doubling of U.S. federal climate research spending (currently about
$4 billion a year) is likely warranted, based on plausible
assumptions about the rate of return on such spending. To avoid
crowd- ing out, this should be phased in to allow a progressive
expansion in supply of college graduates in engineering and
science.
3.2.3 Deployment Policy
In principle there are several possibilities for market failures at
the technology deploy- ment stage. For example, through learning-
by-doing early adopters of a new technology (e.g., a cellulosic
ethanol plant or solar pho- tovoltaic installations) may lower
produc- tion costs for later adopters (e.g., Arthur van Benthem,
Gillingham, and James Sweeney 2008). But, since the potential for
these spill- overs may vary greatly depending on industry
structure, the maturity of the technology, etc., any case for early
adoption subsidies needs to be considered on a case-by-case
basis.
Another possible market failure is con- sumer undervaluation of
energy efficiency,
which has been a key motivation for regu- lations governing auto
fuel economy and household appliances. However, although there is
an empirical literature suggest- ing that households discount
savings from energy efficiency improvements at much higher rates
than market rates, whether this is evidence of a market failure as
opposed to hidden costs or borrowing constraints remains an
unsettled issue (e.g., Gillingham, Newell, and Karen Palmer 2009).
Other mar- ket imperfections might include asymmetric information
between project developers and lenders, network effects in large
integrated systems, and incomplete insurance markets for liability
associated with specific technolo- gies. However, because solid
empirical evi- dence is lacking, little can be said about the
seriousness of all these market failure pos- sibilities, and
whether or not they might war- rant additional policy
interventions.
3.3 International Policy Design
Proposed architectures for international emissions control regimes
can be loosely clas- sified into those based on bottom-up versus
top-down (i.e., internationally negotiated) approaches and
cap-and-trade systems ver- sus systems of emissions taxes (e.g.,
Joseph E. Aldy and Stavins 2007). There is disagree- ment over
which type of architecture is most desirable, and most likely to
emerge in prac- tice. In the bottom up approach, norms for
participation might evolve from small groups of countries launching
regional programs that progressively expand and integrate, or by
explicit linking of domestic cap-and-trade programs (e.g., Carraro
2007; Judson Jaffe and Stavins 2008; Victor 2007). Alternatively,
countries might regularly pledge emissions reductions with periodic
reviews by a for- mal institution (e.g., Thomas Schelling 2007;
Pizer 2007). Here we focus on top-down approaches, given that
advocates of rapid cli- mate stabilization tend to favor
internation- ally binding commitments.
Journal of Economic Literature, Vol. XLVIII (December
2010)926
The most daunting challenge is design- ing an architecture that
encourages par- ticipation among some three or four dozen of the
world’s largest GHG emitters—the Kyoto framework failed to do this
as non- Annex 1 countries, including China, Brazil, South Africa,
Mexico and Indonesia, had no emissions control obligations, while
the United States withdrew from the agree- ment.21 Broad
participation is needed—at least over the longer term and possibly
also the near term under a stringent climate sta- bilization target
(see above)—to promote the cost-effectiveness of any international
agreement, and limit concerns about inter- national competitiveness
and emissions leak- age. Participation of developing countries
through the Clean Development Mechanism (CDM), as at present, does
not reduce global emissions—it only lowers the cost to devel- oped
countries of meeting their emissions goals by allowing firms to
purchase (lower cost) emissions reductions elsewhere on a
project-by-project basis. Moreover, there is considerable concern
that some CDM credits may not represent truly additional
reductions, due the difficulty of establishing a baseline against
which reductions can be measured, in which case the CDM serves to
increase global emissions (e.g., Andrew Keeler and Alexander
Thompson 2008; Rosendahl and Jon Strand 2009).
To be successful, each country must per- ceive an emissions control
agreement as equitable in terms of sharing the burden of global
mitigation costs. Usually this means that industrial countries bear
a dispropor- tionately greater cost burden due to their higher per
capita income and greater con- tribution to historical GHG
accumulations.
21 China’s CO2 emissions n